The tag: URI scheme is a neat little invention by TimKindberg? and SandroHawke that enables people to create simple, fast, and easy identifiers for (often arbitrary) SemanticWeb concepts. Gets rid of the need for persistently maintained Web space, which is cool. More details from:-
http://www.ietf.org/internet-drafts/draft-kindberg-tag-uri-06.txt
Not yet an RFC or anything official, so be warned.
-- SeanPalmer
It's essentially a very simple but effective GloballyUniqueIdentifier algorithm.
It is not a very effective resource-locator algorithm, however, which is often more important than durable/persisent abstract "identity". My biggest problem with tag: URIs ... having just encountered one as the link of a feed-delivered blog post which I now can't navigate to ... is that a browser has no hope of resolving such a tag: URI? At least http: URIs can resolve, and/or redirect, and/or even result in a explicit 404... in the _worst_case_ may resolve to something else entirely, but that's generally obvious. -- JoshSled?
That's not a problem with tag:s, it's a problem with whatever feed you were accessing. I know they're currently used in Atom feeds as a unique identifier for entries, but in whatever situation, a blog feed should always provide a current URL also. -- AdamAtlas?
A TagUri is a UniformResourceIdentifier using the 'tag' scheme; the tag itself follows. The tag starts with an authority identifier, composed of an email address or domain name specific to the authority plus an Iso8601 date (at which that address belonged to the authority; this guards against changes in ownership of addresses), separated by a comma, and followed (after a colon) by an arbitrary name which is unique in the scope of the authority. For example, WardCunningham might choose to identify the Wiki as tag:ward@c2.com,2002-07-26:PprWiki. He might also choose the shorter tag:c2.com,2002:PprWiki.
A possible problem with the system is the use of a date; it means that there isn't a unique, or even obviously canonical, name for an authority, as there may be a range of dates on which the address was bound to that authority. A simple canonicalisation rule might say that the earliest such date should be used, but that particular date might be hard to remember; alternatively, it could be the first instance of some particular recurring date (eg your birthday or the first of January) on which the address was owned, but that isn't general (e.g. if I only have an email address for a week). Of course, if you can store URIs somewhere, so you don't need to regenerate them from scratch, this is moot.
I don't think that a canonical name for an authority is required, the requirement is a unique identifier and thus any date in which the authority had control of the domain works just fine. By default, since -01-01 is implicit when you just provide a year; most people will use the closest year in which they were the holder of the domain on January 1st. -- ClarkEvans
TagUris are used to identify types in YamlAintMarkupLanguage.